User Trajectory Pattern Change Analysis Based On AP Logs

With the rapid development and popularization of WIFI positioning technology,a large number of location and trajectory data are recorded in AP logs.Behind these trajectory information,there are rich user movement characteristics and regulations.In this paper,we aim to find effective ways to mine user mobility patterns from the massive AP logs,analyzes mobility pattern change through corresponding algorithms,and applies it to position prediction.Firstly,this paper analyzes the existing sequence pattern mining algorithms based on adaptive sliding time window and improved PrefixSpan,and to improve on its shortcomings,designed an incremental sequence pattern mining algorithm based on an adaptive sliding window to extract the trajectory patterns in the time window,that is,to add new sequence pattern or adjust the sequence support,only need to dynamically adjust the current sequence pattern.Through experimental comparison,it is found that the algorithm improves the efficiency of sequential pattern mining.Secondly,this paper deeply studies the trajectory pattern change analysis algorithm,which defines a set of indicators that can accurately quantify the trajectory pattern changes,and finds the changing time point of the trajectory pattern through Bayesian analysis.The algorithm was simulated and analyzed using the collected AP log data.The results show that the algorithm's quantization method truly reflects the change of the user's trajectory pattern and effectively detects the trajectory pattern change time point.In addition,this paper also studies the influence of external events on the change of user trajectory pattern.The results show that the changes of the group trajectory pattern are closely related to external group events such as holidays.Finally,this paper applies the trajectory pattern change analysis algorithm to the user position prediction,and designs a sub-trajectory synthesis position prediction algorithm based on pattern change and similar patterns.The experimental comparison shows that the sub-trajectory synthesis algorithm based on pattern change can find the trajectory patternchange in time,and adopt the historical trajectory of the previous pattern change point so far as the training dataset,which improves the efficiency and accuracy of the algorithm;The sub-trajectory synthesis algorithm based on similar patterns adds similar users for auxiliary prediction,which improves the accuracy of the algorithm.